Can We Automate Scientific Reasoning in Closed-Loop Experiments using Large Language Models?

Abdoulatif Cissé, Max E. Cooper, Mengjia Zhu, Xenophon Evangelopoulos, Andrew I. Cooper

Submission Type: I want my submission to be considered for both oral and poster presentation.

Keywords: Large Language Models (LLMs); scientific reasoning; automated experimentation; Bayesian optimisation (BO)

TL;DR: We systematically compare BO, hybrid LLM+BO, and LLM-only methods in closed-loop experiments, finding that LLM-only approaches often match or outperform alternatives under fixed budgets, while exhibiting problem-dependent strengths and limitations.

Confirmation Of Submission Requirements: I submit a previously published paper. It was published in an archival peer–reviewed venue on or after September 1st 2025, I specify the DOI in the field below, and I submit the camera-ready version of the paper.

DOI: https://doi.org/10.1039/D5DD00520E

Submission Number: 335